It has been recognized that CORSIM (and its constituent program, NETSIM) is one of the most widely used and effective computer programs for the simulation of traffic behavior on urban transportation networks. Its popularity is due in large part to the high level of detail incorporated into its modeling routines. However, the car-following models, used for the simulation of driver behavior in the program, have not been formally calibrated or validated. Since the model has performed well in a wide range of applications for so many years, it has always been assumed to have an implied validity. This study evaluated the NETSIM car-following models by comparing their results with field data. Car-following field data were collected using a new data collection system that incorporates new Global Positioning System and geographic information system technologies to improve the accuracy, ease, speed, and cost-effectiveness of car-following data collection activities. First, vehicle position and speed characteristics were collected under field conditions. Then simulated speeds and distances were based on identical lead vehicle actions using NETSIM car-following equations. Comparisons of simulated and field data were completed using both graphical and statistical methods. Although some differences were evident in the graphical comparisons, the graphs overall indicated a reasonable match between the field and simulated vehicle movements. Three statistical tests, including a goodness-of-fit test, appear to support these subjective conclusions. However, it was also found that definitive statistical conclusions were difficult to draw since no single test was able to compare the sets of speed and distance information on a truly impartial basis.
An autonomous navigational system using GPS and computer vision for futuristic road traffic
